v1.0.0

Trimmer High

BenignClawScan verdict for this skill. Analyzed Apr 30, 2026, 10:43 PM.

Analysis

This instruction-only skill is coherent for cloud video trimming, but it sends videos and prompts to a NemoVideo API and uses a bearer token-backed session for uploads, edits, and exports.

GuidanceBefore installing, be comfortable sending raw videos and editing prompts to https://mega-api-prod.nemovideo.ai. Keep NEMO_TOKEN private, consider using the anonymous limited token, and review export or credit-using actions for important projects.

Findings (8)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

Abnormal behavior control

Checks for instructions or behavior that redirect the agent, misuse tools, execute unexpected code, cascade across systems, exploit user trust, or continue outside the intended task.

Agent Goal Hijack

SeverityLowConfidenceMediumStatusNote

SKILL.md

Backend says | You do ... "click [button]" / "点击" | Execute via API ... "Export button" / "导出" | Execute export workflow

The skill makes backend text authoritative for follow-on API actions. This is aligned with translating a GUI-based video backend into API calls, but users should know remote responses can drive edits or exports within the session.

User impactA backend response can cause the agent to take additional video-editing or export actions instead of only displaying the response to the user.

RecommendationUse it when you are comfortable with backend-guided automation, and ask the agent to confirm before exporting or spending credits if that matters to you.

Tool Misuse and Exploitation

SeverityMediumConfidenceHighStatusNote

SKILL.md

Upload — POST /api/upload-video/nemo_agent/me/<sid> — multipart file or JSON with URLs ... Export — POST /api/render/proxy/lambda with render ID and draft JSON. Poll GET ... every 30s

The skill exposes upload, render, polling, state, and export API operations. These are expected for cloud video trimming, but they can transmit user media and consume service credits.

User impactYour uploaded footage and editing requests are processed by the external service, and export actions may use the service account or credit balance tied to the token.

RecommendationOnly upload footage you are willing to process in the NemoVideo cloud, and review export/credit actions when using paid or limited tokens.

Agentic Supply Chain Vulnerabilities

SeverityLowConfidenceMediumStatusNote

metadata

Source: unknown; Homepage: none; No install spec — this is an instruction-only skill.

The skill has no local install mechanism or code, which limits local supply-chain exposure, but its origin and project homepage are not identified.

User impactIt may be harder to independently verify who maintains the skill or the external service it relies on.

RecommendationPrefer limited tokens and avoid uploading highly sensitive footage unless you trust the skill owner and the NemoVideo service.

Cascading Failures

SeverityLowConfidenceMediumStatusNote

SKILL.md

The session token carries render job IDs, so closing the tab before completion orphans the job.

The cloud render job can continue or become detached from the user session if interrupted. This is disclosed and tied to the render workflow, but it is a containment/recovery consideration.

User impactAn interrupted export may leave a cloud job running or difficult to recover, potentially wasting time or credits.

RecommendationKeep the session open until exports complete, and check credit balance or job status after interrupted renders.

Rogue Agents

SeverityLowConfidenceMediumStatusNote

SKILL.md

closing the tab before completion orphans the job

The artifact discloses that a cloud render job may continue outside the visible chat/session flow. This is part of the intended rendering pipeline, not hidden persistence or self-propagation.

User impactA render job may keep running after the user interface is closed, even though no local background process is shown.

RecommendationAvoid starting exports you do not want completed, and monitor cloud job status when closing or interrupting the session.

Permission boundary

Checks whether tool use, credentials, dependencies, identity, account access, or inter-agent boundaries are broader than the stated purpose.

Identity and Privilege Abuse

SeverityLowConfidenceHighStatusNote

SKILL.md

All requests must include: Authorization: Bearer <NEMO_TOKEN> ... Don't print tokens or raw JSON.

The skill requires a bearer token for the video-processing API and explicitly tells the agent not to print it. This credential use is expected for the stated service integration.

User impactAnyone with the token may be able to use the associated NemoVideo API access or credits.

RecommendationKeep NEMO_TOKEN private, rotate it if exposed, and use an anonymous or limited token where possible.

Sensitive data protection

Checks for exposed credentials, poisoned memory or context, unclear communication boundaries, or sensitive data that could leave the user's control.

Memory and Context Poisoning

SeverityMediumConfidenceHighStatusNote

SKILL.md

Save `session_id` from the response ... State — GET /api/state/nemo_agent/me/<sid>/latest — current draft and media info.

The skill stores and reuses a cloud session identifier and retrieves draft/media state from the backend. This is necessary for editing continuity, but it means session context and media metadata exist outside the local chat.

User impactYour project state, media references, and render/job context may remain associated with the cloud session while the workflow is active.

RecommendationAvoid uploading confidential footage unless you are comfortable with cloud-side session state, and start a new session for unrelated projects.

Insecure Inter-Agent Communication

SeverityMediumConfidenceHighStatusNote

SKILL.md

Chat (SSE) — POST /run_sse with `session_id` and your message in `new_message.parts[0].text` ... Tool call/result | Process internally, don't forward

The skill communicates with a backend agent/service over SSE and sends user messages plus session identifiers. The destination is disclosed, but this is still an external agent/provider boundary.

User impactPrompts, editing instructions, and session data are sent to the NemoVideo backend, and backend tool activity may be summarized rather than shown raw.

RecommendationDo not include unrelated secrets in editing prompts, and use the skill only if you are comfortable with the disclosed provider receiving the video-editing context.